Project 3
Karl Jackson
2025-04-30
Introduction
The two words I chose to use to search for photos on pexels.com were “crispy” and “potatoes”, this is due entirely to the fact that I was hungry when I started the project. Another reason is also is that crispy potatoes are my favorite food and I couldn’t think of anything else to use for images
- most of the photos aren’t actually about crispy potatoes but are about meals with crispy potatoes as a side dish
- most of the photos were about fries instead of crispy potatoes
- most of the photos had the fries (crispy potatoes) off to the side and out of focus
knitr::kable(select(selected_photos, url))Key features of my selected photos
- The mean display ratio of the images is 0.9820233, which means that most images are almost a square
- The mean width of each images is 4274.9047619, so each image is around 4000x4000 pixels
- The mean percentage that the highest color takes up in the image is 0.4839299, so most images are around 48% red
- The mean alt text length is 79.9047619, so each image takes about 80 words to describe
Creativity
I wanted to make an anti-meme gif, so I thought it would be funny If it was just a gif of images of potatoes that had the word potato in the middle
I used some basic loops and operations to make the text colored, and to split the word “potato” into 3 parts “po”, “ta”, and “to”
I did this mainly because I thought it would make a better anti-meme
Learning reflection
I liked learning the basics of accessing API’s in r, but I would of liked to have done a more in-depth project with a much larger API
I think something interesting to explore would be using the statistics features in R in combination with API’s to explore large data sets and gather information about them. One of the biggest reasons why I am interested in this is because I don’t think R has any advantages over other languages in accessing API’s in the conventional sense, python can already do pretty much everything we have done in this project but I think R would most likely be better equipped to analyzing a large API if we were examining if for a statistical purpose instead of creating a meme
Appendix
library(tidyverse)
library(httr)
library(magick)
# function to find the most common color from a hex string
most_color <- function(hex_str) {
# removes the # at the start of a hex string
hex_str <- substr(hex_str, 2, 7)
# getting the parts of the hex string
str_red <- substr(hex_str, 1, 2)
str_blue <- substr(hex_str, 3, 4)
str_green <- substr(hex_str, 5, 6)
# turning the parts into base 10
red <- as.numeric(as.hexmode(str_red))
blue <- as.numeric(as.hexmode(str_blue))
green <- as.numeric(as.hexmode(str_green))
# getting the most common color and the position to return the color
colors <- c(red, blue, green)
most_color <- max(colors)
return (most_color / sum(colors))
}
api_key <- "2Mqpbjk1WNNf3bsNN5cU67pthlZPNBi5MiKs8uUnwscAlQ1FpNrL2zXb"
url <- "https://api.pexels.com/v1/search?query=crispy%20potatoes&per_page=80"
response <- httr::GET(url,
add_headers(Authorization = api_key))
data <- httr::content(response,
as = "parsed",
type = "application/json")
photo_data <- tibble(photos = data$photos) %>%
unnest_wider(photos) %>%
unnest_wider(src)
# variable 1 - display ratio (width / height)
# variable 2 - if the alt string (describes the image) has the word fries
# variable 3 - the percentage of the average color that belongs to the most common color
# filters based on images with an average color that is at least 44% the most common color
selected_photos <- photo_data %>%
group_by(avg_color) %>%
mutate(
display_ratio = width / height,
alt_has_fries = str_detect(str_to_lower(alt), "fries"),
alt_length = str_count(alt),
most_color = most_color(avg_color)
) %>%
filter(most_color > 0.44)
write_csv(selected_photos, "selected_photos.csv")
mean_ratio <- selected_photos$display_ratio %>% mean(na.rm = TRUE)
mean_color_percent <- selected_photos$most_color %>% mean(na.rm = TRUE)
mean_width <- selected_photos$width %>% mean(na.rm = TRUE)
# grouping the photos by which ones have "fries" and then finding the mean number of words in the alt string
grouped_photos <- selected_photos %>% group_by(alt_has_fries)
grouped_photos %>% summarise(avg_words = mean(alt_length))
urls <- selected_photos$medium
# setting up some vectors for the animation
colors = c("red", "orange", "yellow", "green", "blue", "purple", "violet")
potato <- c("PO", "TA", "TO")
images <- c()
for (i in 1:length(urls)) {
# makes all images a constant size for the gif
image <- image_read(urls[i]) %>% image_scale("850") %>% image_crop("500x500")
# this adds a white "border" around the text as when the text is blue its sometimes hard to see
image <- image %>% image_annotate(potato[(i %% length(potato) + 1)], size=205, gravity="Center", color="white")
# adding the text to the image to make it say po -> ta -> to
image <- image %>% image_annotate(potato[(i %% length(potato) + 1)], size=200, gravity="Center", color=colors[(i %% length(colors)) + 1])
images <- image %>% append(x=images)
}
image_animate(images, fps=2, optimize=TRUE) %>% image_write("creativity.gif")---
title: "Project 3"
author: "Karl Jackson"
output: html_document
date: "2025-04-30"
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo=TRUE, message=FALSE, warning=FALSE, error=FALSE)
library(tidyverse)
selected_photos <- read_csv("selected_photos.csv")
```
```{css echo=FALSE}
body {
background-color: black;
color: white;
}
```
## Introduction
The two words I chose to use to search for photos on pexels.com were "crispy" and "potatoes", this is due entirely to the fact that I was hungry when I started the project. Another reason is also is that crispy potatoes are my favorite food and I couldn't think of anything else to use for images
([crispy potatoes](https://www.pexels.com/search/crispy%20potatoes/))
1. most of the photos aren't actually about crispy potatoes but are about meals with crispy potatoes as a side dish
2. most of the photos were about fries instead of crispy potatoes
3. most of the photos had the fries (crispy potatoes) off to the side and out of focus
```{r}
knitr::kable(select(selected_photos, url))
```
## Key features of my selected photos
```{r include=FALSE}
mean_ratio <- selected_photos$display_ratio %>% mean(na.rm = TRUE)
mean_color_percent <- selected_photos$most_color %>% mean(na.rm = TRUE)
mean_width <- selected_photos$width %>% mean(na.rm = TRUE)
mean_text_length <- selected_photos$alt_length %>% mean(na.rm = TRUE)
```
1. The mean display ratio of the images is `r mean_ratio`, which means that most images are almost a square
2. The mean width of each images is `r mean_width`, so each image is around 4000x4000 pixels
3. The mean percentage that the highest color takes up in the image is `r mean_color_percent`, so most images are around 48% red
4. The mean alt text length is `r mean_text_length`, so each image takes about 80 words to describe
## Creativity
I wanted to make an anti-meme gif, so I thought it would be funny If it was just a gif of images of potatoes that had the word potato in the middle
I used some basic loops and operations to make the text colored, and to split the word "potato" into 3 parts "po", "ta", and "to"
I did this mainly because I thought it would make a better anti-meme
## Learning reflection
I liked learning the basics of accessing API's in r, but I would of liked to have done a more in-depth project with a much larger API
I think something interesting to explore would be using the statistics features in R in combination with API's to explore large data sets and gather information about them. One of the biggest reasons why I am interested in this is because I don't think R has any advantages over other languages in accessing API's in the conventional sense, python can already do pretty much everything we have done in this project but I think R would most likely be better equipped to analyzing a large API if we were examining if for a statistical purpose instead of creating a meme
## Appendix
```{r file='exploration.R', eval=FALSE, echo=TRUE}
```
```{r file='project3_report.Rmd', eval=FALSE, echo=TRUE}
```